Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!zaphod.mps.ohio-state.edu!pacific.mps.ohio-state.edu!linac!att!pacbell.com!ames!sgi!shinobu!fido.wpd.sgi.com!bh
From: bh@sgi.com (Bent Hagemark)
Newsgroups: comp.sys.sgi
Subject: Re: R2000 stack alignment question
Message-ID: <1991May30.200122.15985@fido.wpd.sgi.com>
Date: 30 May 91 20:01:22 GMT
References: <9105281723.AA11513@crow.omni.co>
Sender: news@fido.wpd.sgi.com (Usenet News Admin)
Organization: Silicon Graphics, Inc., Mountain View, CA
Lines: 25

In article <9105281723.AA11513@crow.omni.co> rpaul@crow.UUCP (Rodian Paul) writes:
>> After some hours of digging through assembly code I've finally found out
>> that varargs.h depends on the stack being aligned to an 8-byte boundary,
>> and the C compiler does its utmost to keep it like that (unlike me).
>> 
>Don't use varargs(5) it is non-standard, use stdarg(5) instead if you want
>portabilty via ANSI C.

Absolutely (use stdarg(5)).  The following is from the 4.0 stdarg(5) man page:

     The va_start (ap, ParmN) macro initializes ap for subsequent use by
     va_arg and va_end.  va_start must be called before any use of va_arg.
*    The ANSI C Standard (ANSI X3.159-1989) restricts the type of parmN to one
*    of the types resulting from the default argument promotions, currently
*    int, unsigned int, pointer, or double.  If va_start is invoked with a
*    parmN which is not one of these types (e.g., if parmN is short or char)
*    the behavior is undefined.

After learing this the hard way I asked that the *'ed lines be added
to the stdarg(5) man page.

The default argument promotion rules are described in section 2.7
of K&R C, 2nd Edition.

Bent